Listen to your machine with your phone and catch when it drifts from “its normal”.
This page is both a user guide and a small vibration/sound-analysis handbook — you can learn it from scratch.
Check It is not a fault-diagnosis device. It is a change-tracking and triage tool. In one sentence:
To do that it uses two of your phone’s sensors:
On each listening it extracts a sound/vibration fingerprint and compares it to the learned “normal”. If the difference is large, it warns you. The core listening and analysis happen on your phone; the optional Oney AI advisor (PRO) is the one feature that uses the internet, and only with your consent.
Give it a name (e.g. “AC”, “Pump 3”), pick its type (motor/pump/fan/fridge/car…). The app auto-selects the right sensor for that type and honestly tells you how much it can catch on that machine. If you know the speed (RPM), enter it — for the fault-type hint.
While the machine runs normally, rest the phone firmly on its housing for ~20 s. The app learns its own normal. Important: don’t record on a stopped/faulty machine — “normal” would be learned wrong.
Later, rest the phone again → it shows 🟢🟡🔴 + a score + a fault-type hint + a trend. If you listen regularly (e.g. weekly), the trend becomes meaningful.
Every listening is saved. Tap the machine → past records + trend graph: see whether it is degrading over time.
| Indicator | Meaning |
|---|---|
| 🟢 HEALTHY | Low drift from normal — the machine is close to how it was learned. |
| 🟡 WATCH | Noticeable drift — track it, re-measure, check conditions. |
| 🔴 CAUTION | Large drift — something changed. Have an expert verify. |
| Score (0–100) | Closer to 100 = closer to normal. Not an absolute “health grade”; it’s a fit to its own normal. |
| HI | Health Index — the raw distance from normal (compared against a threshold τ). |
| 🔎 hint | A likely fault-type guess from the spectrum shape (not certain — see below). |
This part helps you understand the graphs and “hints” in the app. You can follow it even with no prior knowledge.
A vibration/sound signal rises and falls over time (a wave). If we split the same signal into “which frequencies it is made of” using the FFT (Fast Fourier Transform), we get its spectrum. Fault signatures hide in specific peaks of the spectrum.
In the app you see both graphs live: wave + spectrum (bars).
A rotating machine’s base frequency is 1× = RPM ÷ 60 (Hz). Example: 1797 RPM → 1× ≈ 29.95 Hz. Its multiples 2×, 3×… are harmonics. Which harmonic dominates suggests a likely fault type:
| What you see in the spectrum | Likely type (hint) |
|---|---|
| 1× dominant, few harmonics | Imbalance / unbalance |
| 2× high (more than half of 1×) | Misalignment (coupling) |
| Many harmonics (1×+2×+3×+…) | Mechanical looseness |
| Broadband high frequency ↑ | Cavitation / wear / rubbing |
Bearing faults produce their own characteristic frequencies. These are computed from the bearing geometry and speed by physics (not learned from example data):
| Abbr. | What |
|---|---|
| BPFO | Ball Pass Frequency, Outer race |
| BPFI | Ball Pass Frequency, Inner race |
| BSF | Ball/roller Spin Frequency |
| FTF | Fundamental Train (cage) Frequency |
In the Pro section, if you enter the bearing geometry (ball count N, ball diameter d, pitch diameter D), it computes these frequencies exactly and checks whether there is a prominent peak at that point in the spectrum. Important limit: early bearing faults are high-frequency; a phone accelerometer is weak there → for bearings the phone is limited (let’s be honest).
For vibration velocity (mm/s RMS, 10–1000 Hz) there are international severity zones:
| Zone | Meaning |
|---|---|
| A / B | Good / acceptable (newly commissioned, or fine for long-term running) |
| C | Borderline — not satisfactory for long-term, plan action |
| D | Can cause damage — take action |
The Pro section computes the ISO zone when the unit is mm/s and you selected the machine group + mount type. If the unit is g / m/s² (raw phone acceleration) ISO is not given directly (honest: velocity integration is a separate step) — instead a band-RMS is shown for information.
On phones that have an accelerometer, for vibration machines Check It combines microphone + accelerometer (fusion). Why? Because the accelerometer is immune to acoustic noise: even if the surroundings are noisy, the vibration channel stays clean. In our lab measurements, for low-frequency faults (imbalance/misalignment) fusion was noticeably more robust under noise than mic-only.
The Pro section loads an external vibration CSV or sound WAV and analyses it in detail (ideal if you captured data from a dedicated vibration device).
Listen live with the mic amplified (use headphones). Like a digital stethoscope — catch sounds you can’t hear.
Record + label your own sounds → the app recognises them live. Plus 521 built-in sounds (dog/bird/car…) recognised on-device (YAMNet).
Measure several units → it finds the one most deviating from the average. Quick triage for a fleet/workshop.
Compare two measurements (before↔after maintenance, last month↔this month): % change + difference spectrum.
As you do regular checks you earn 🌱→🥉→🥈→🏆 badges — to build the habit.
Not for the core change-tracking — it runs fully on your device. The optional Oney AI advisor (PRO) does need internet (it sends your question to a cloud AI provider, only with your consent).
With no accelerometer it works mic-only (it says so honestly). Most phones have no ambient-temperature sensor; if present it uses it, otherwise it says “none” (we never show battery heat as machine heat).
On the main screen: Backup (a single JSON file) / Restore. Your data is yours; export/delete anytime.
The score is relative to the normal you learned. If you taught the normal wrong (while faulty), it misleads — teach it while healthy.